The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages a vendor-specific API key (PIXAZO_API_KEY). It correctly recommends storing the key in environment variables or a dedicated local file (~/.pixazo/api-key) with restricted permissions (chmod 600), which is a recognized safe practice for secret management.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection as it interpolates untrusted user input (text prompts and reference image URLs) into API requests. However, the skill explicitly includes a safety mechanism (enable_safety_checker: true) in its payload to mitigate the processing of malicious or unsafe content.
Ingestion points: User-provided text prompts and reference image data/URLs in SKILL.md.
Boundary markers: The skill uses structured JSON payloads for API communication, which provides structural boundaries for the input data.
Capability inventory: The skill performs network POST requests to gateway.pixazo.ai and writes to the local file system at ~/.pixazo/api-key.
Sanitization: The API request includes an enable_safety_checker parameter set to true to filter output and potentially input at the model level.
[EXTERNAL_DOWNLOADS]: The skill references external resources including documentation from www.pixazo.ai and a command to add additional skills via npx from the Pixazo-AI GitHub organization. These resources are managed by the skill's vendor and represent standard operational behavior.

ernie-image